Enhancing Early Detection of Heart Disease Using Ensemble, Traditional, and Hybrid Machine Learning Techniques

Early detection of heart disease is critical for effective intervention, yet conventional diagnostic methods are limited by subjectivity and scalability. This paper investigates machine learning (ML) approaches for heart disease prediction using the Cleveland dataset comprising 303 patient records and 13 clinical features. Ten models were evaluated, including Logistic Regression, Naïve Bayes, Support Vector Machine, Random Forest, XGBoost, a neural network, and two hybrid frameworks. The proposed Hybrid XGGA integrates eXtreme Gradient Boosting with Giant Armadillo Optimization for feature selection, while the Hybrid XAI-ML combines ensemble learning with SHAP- and LIME-based interpretability. Experimental evaluation using accuracy, precision, recall, F1score, and ROC-AUC demonstrates that Hybrid XAI-ML achieved the best performance (96.17 % accuracy), followed by Hybrid XGGA (95.26 %). Computational analysis confirmed their practicality with moderate training overhead. The results highlight that integrating optimization and explainable AI enhances both predictive accuracy and interpretability, supporting the development of transparent decision-support systems for cardiovascular risk assessment.